wxauto-mcp has 19 tools with descriptions in Chinese covering WeChat automation. Tool definitions are directly visible in the source code. However, critical quality gaps limit the score: (1) Descriptions are terse and lack structured context on when/why to use tools. Most are 20-80 chars, well below the 50-200 char production baseline. (2) Parameter descriptions are minimal; many lack type clarity or constraints. (3) No output schemas are documented, responses are completely opaque to the LLM. (4) No error handling guidance, tools provide no recovery hints if they fail. (5) Security concerns: wechat_initialize and wechat_activate accept credentials as parameters (license_key), violating secret-injection patterns. (6) Naming follows verb_noun convention (good), but many parameter names lack semantic clarity (e.g., 'who' is ambiguous, is it a username, phone, or internal ID?). Overall, the tools are functional but underdocumented for agentic use, an LLM would struggle to understand when to call send_message vs send_files vs send_url_card without deeper reasoning.

• Translate all tool descriptions to English and expand from terse Chinese to 50-200 chars covering WHAT the tool does, WHEN to use it, and key prerequisites. Example: 'send_message' → 'Send a text message to a WeChat contact. Use this after switch_chat() to send to a specific person, or omit 'who' to send to the current active window. Returns success status and message ID.'
• Document output schema for every tool. Return examples like: get_messages → {messages: [{id: string, from: string, content: string, timestamp: ISO8601}], total: integer, has_more: boolean}. Include field descriptions and types.
• Replace license_key parameter with server-side environment variable injection. Store WXAUTO_LICENSE_KEY in .env, read at startup in wechat_initialize(). Remove license_key from wechat_activate(), activate on first initialize(). Document in README.
• Clarify parameter 'who' across all tools: rename to 'contact_name' or 'recipient_name' and document expected format (e.g., 'Display name as shown in contacts; if multiple matches, the first alphabetically is used'). Alternatively, provide separate variants: send_message_by_name() and send_message_by_id().
• Add bounds to numeric parameters: get_history(n: 1-500, default 50), get_friends(n: 1-200, default 10), send_url_card(timeout: 1-300 seconds, default 10), add_friend(timeout: 1-60 seconds, default 5).
• Add error handling guidance to descriptions of high-risk tools: send_message → 'Returns {success: true, message_id: string} on success. On error (contact not found, send timeout), returns {success: false, error: string, recovery_hint: string}. Call switch_chat() first to verify contact exists.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry